Electronic and electrical appliance control method and control equipment based on reinforcement learning, and storage medium

Abstract

The invention relates to an electronic appliance control method based on reinforcement learning. For an electronic appliance with an automatic scene control function, reinforcement learning control policies are applied, intervention control of a user on electronic and electrical equipment continuously is acquired to serve as decision input of reinforcement learning, a scene algorithm model adapting to automatic control of equipment under different scenes of a user is dynamically generated, a scene algorithm model of an automatic working mode of the electronic appliance closest to the use habitof the user is obtained, and the use efficiency of the electronic appliance is improved. The design method is applied to various electronic appliances, so that optimization and updating of a self-learning mode of automatic control of all-electronic appliance scenes are realized, and a better automatic scene control method is provided for smart homes and smart offices.

Description

technical field [0001] The invention relates to a control method for electronic appliances based on reinforcement learning, a control device, and a storage medium, and belongs to the technical field of building IoT intelligence. Background technique [0002] At present, most of the smart home systems on the market rely on the two functions of "scene" and "automation" to complete most of the functions, and the control method mainly relies on voice control or mobile phone control. Although many users feel that artificial intelligence is very developed now, the system should be able to learn user habits by itself, and be able to automatically apply scenarios centered on the home, home devices and the outside world for interaction and feedback (through device monitoring and automatic linkage with scene devices, it can collect personal or environmental information ), thereby greatly improving the comfort of home life, but in fact, the current application of AI in the field of sma...

AI Technical Summary

Problems solved by technology

In general, an action will not only affect the current reward, but also affect the environment at the next time point, so it will also affect all subsequent rewards

Because the actions of the learning system will affect the environment, and the environment will affect the subsequent actions, so in essence, reinforcement learning is a closed-loop control problem

Images